JPH07320496A - Non-volatile semiconductor memory - Google Patents

Abstract

PURPOSE:To obtain a nonvolatile semiconductor memory which can relieve a cell having defective insulating film and of defective write-in, simplify the erasing sequence and shorten the time for it. CONSTITUTION:This device is a flash memory provided with a redundancy circuit relieving a defective cell by replacing it with a spare cell. A row address of a defective cell to be replaced with a spare cell is stored in a redundancy circuit ROM 29, it is monitored by a comparator 26 whether an inputted row address coincides with a defective row address stored in the redundancy ROM 29 or not, and when they coincide with each other, the row address stored in the redundancy ROM 29 is selected by an address multiplexer 22, and transferred to an address bus AB. When a negative potential is applied to a control gate of a cell and erasing is performed, since a defective row is selected and fixed to a ground potential, it can be evaded that a negative potential is applied to a defective row, and a cell of a defective insulating film and a cell of defective write-in can be surely relieved.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a non-volatile semiconductor memory device, and more particularly to a row decode circuit in a flash memory.

[0002]

2. Description of the Related Art A flash memory is a nonvolatile semiconductor memory device (EEPROM: Electrically Erasable Program) capable of electrically writing and erasing data.
mmableRead Only Memory). As shown in FIG. 10, a flash memory cell, which is currently in the mainstream, has a double-layer gate structure EPROM (Erasable Read Only Mem).
ory: It has the same structure as an ultraviolet erasable EPROM). In FIG. 10, 11 is a semiconductor substrate, 12 is a source, 13 is a drain, 14 is a tunnel oxide film, 15 is a floating gate, 16 is an insulating film (called an interpoly insulating film), and 17 is a control gate. A source 12 and a drain 13 are formed separately in a semiconductor substrate 11, and a tunnel insulating film 14, a floating gate 15, an interpoly insulating film 16 and a control gate 17 are formed on a channel region between the source 12 and the drain 13. The cells are stacked to form a cell.

To write data in the flash memory, the source 12 is grounded, and high potentials for writing are applied to the control gate 17 and the drain 13, respectively, in the same manner as in EPROM, so that hot electrons generated near the drain 13 are generated. It is performed by injecting into the floating gate 15 and raising the threshold voltage of the cell transistor.

On the other hand, a general erasing method is shown in FIG.
, The control gate 17 is grounded, the drain 13 is open, and the source 12 has a high potential (Vs1) for erasing.
Is applied, a high electric field is applied between the source 12 and the floating gate 15 to generate a tunnel current, and the electrons in the floating gate 15 are extracted to the source 12 and a source erasing method (hereinafter abbreviated as SE method) is adopted. There is.

However, in the case of the SE method, it is necessary to increase the electric field strength between the source 12 and the floating gate 15 until a sufficient tunnel current can be generated, and if the thickness of the tunnel oxide film 14 is set to about 10 nm, the source A high potential of 10 V or higher must be applied to 12. In order to apply a high potential of 10 V or higher to the source 12, it is necessary to increase the breakdown voltage of the source 12. Therefore, it is necessary to make the impurity concentration profile have a double structure as shown in FIG. 12, that is, to cover the source 12 with the low-concentration impurity diffusion layer 18, and the source region must be wide. Adopting such a configuration is disadvantageous in terms of cell miniaturization, which is one of the most important requirements in memory. Note that in FIG. 12, two adjacent memory cells are extracted and shown.

Therefore, in order to avoid the problems caused by the SE method as described above, the following erasing method has been proposed. In this method, as shown in FIG. 13, a high potential (Vs2) for erasing is applied to the source 12, the drain 13 is opened, and a negative potential Vg is applied to the control gate 17, and the source gate erasing (hereinafter SGE) is performed. (Abbreviated as method)
It is called. In the SGE method, the electric field strength between the source 12 and the floating gate 15 necessary for causing the tunnel phenomenon is obtained by biasing the control gate 17 with a negative potential. This makes Source 1
The high potential Vs2 applied to 2 is lower than the high potential Vs1 of the SE method (Vs2 <Vs1), and the high breakdown voltage design of the source 12 as shown in FIG. No need. Therefore, the control gate 17,
In other words, although it is necessary to apply a negative potential to the word line,
It is more advantageous than the SE method in terms of cell miniaturization. Furthermore,
In both of the SE and SGE methods, an erase current larger than that of the control gate 17 flows through the source 12, but in the SGE method, the bias level of the source 12 that requires current supply can be lowered, so that the write The power supply voltage Vpp can be generated by the booster circuit provided inside the chip, and a single power supply (normal power supply: Vcc) is possible. Therefore, it can be said that the SGE method is also excellent in this respect. In both methods, the electrons in the floating gate 15 are extracted and the threshold voltage of the cell transistor is lowered to erase the data.

Next, a row decode circuit for realizing the above-mentioned SGE system will be described. FIG. 14 shows the row decode circuit in the flash memory and the peripheral circuit section related to the erase operation in an extracted manner. The address signal Add supplied from the outside is input to the address buffer 21, and the output of the address buffer 21 is supplied to the address multiplexer 22 and the address latch circuit 23. The output of the address latch circuit 23 and the output of the address counter 24 are respectively supplied to the address multiplexer 22, and any one of the output of the address buffer 21, the address latch circuit 23 and the address counter 24 is selected, and the internal address bus AB is selected. It is supplied to the row predecoder 25 and the comparator 26 via the.
The row predecoder 25 includes AND gates 27, 27,
Of the address signals supplied via the internal address bus AB, and a row address signal RAdd is supplied to each of them. Also, each AND gate 2
The activation signal PRE of the row predecoder 25 output from the NOR gate 28 is supplied to 7, 27, ...

The comparator 26 compares the address supplied via the internal address bus AB with the defective address stored in the redundancy ROM 29, and outputs a match signal HIT when they match. Redundancy RO
M29 stores the address of the defective cell, and this ROM 29 can store as many bits as the number of row addresses in the case of row redundancy.
On the other hand, if each word line is replaced, all row addresses are stored in the redundancy ROM 29.
Further, if the n-th power of 2 is collectively replaced, such as 2 rows or 4 rows, the address to be stored is reduced by n bits. For this storage, writing and erasing of data may be performed using a flash memory cell, or a method of storing a defective address by providing a fuse using polysilicon and fusing with a laser may be adopted. .

In this way, the defective address is stored in the redundancy ROM 29, and the comparator 26 always checks whether the selected address matches the defective address. When the selected address matches the defective address, the match signal HIT becomes "H" level. As a result, the activation signal PRE becomes "L" level, the predecoder 25 is deactivated, the defective row is deselected, and the level shifter 3 in the spare row decoder 35 is inactivated.
6, the spare word line SWL is driven through the buffer 6 and the buffer 37, and replacement with the spare cell is performed. These replacements are similar to EPROMs. At the time of erasing, since the word lines WL are collectively operated and all are in the non-selected state and are collectively erased including the spare row, the control by the redundancy is not particularly performed.

The coincidence signal H is applied to the NOR gate 28.
IT and the erase signal ERS are supplied, and the output of the row predecoder 25 is prohibited (fixed to “L” level) when the input address and the defective address match each other and in the erased state.

The row predecode signal RPD output from the row predecoder 25 is supplied to the main decoder 30. The main decoder 30 includes an AND gate 3 corresponding to each word line WL in the memory cell array 31.
2, 32, ..., Level shifter 33, ... And buffer 3
4, ... are provided. The level shifter 33 is a circuit that converts a signal level in order to bring the word line WL to a high potential during writing, and converts a Vcc system signal into a Vpr system signal and outputs it.

The level shifter 33 includes P-channel MOS transistors Q1 and Q as shown in FIG.
2, N-channel type MOS transistors Q3 and Q4, and an inverter 39. The potential Vpr is applied to the sources of the MOS transistors Q1 and Q2,
A MOS transistor Q is connected between each drain and ground point GND.
The drains and sources of Q3 and Q4 are connected. The gate of the MOS transistor Q1 is the above-mentioned MOS transistor Q.
2, the drain of Q4 is connected to the common connection point, and the gate of the MOS transistor Q2 is the above-mentioned MOS transistor Q1,
It is connected to the common drain connection point of Q3. The output signal of the AND gate 32 is supplied to the gate of the MOS transistor Q3, and the output signal of the AND gate 32 is supplied to the gate of the MOS transistor Q4 via the inverter 39. The output signal obtained from the connection point of the MOS transistors Q1 and Q3 is supplied to the buffer 34.

The buffer 34 includes a P-channel type MOS transistor Q5 and an N-channel type M as shown in FIG.
It is composed of a CMOS inverter including an OS transistor Q6. The operating power supply of this CMOS inverter is the potential Vpr and the bias potential Vbb, and the output of this buffer 34 drives the corresponding word line WL in the memory cell array 31.

The memory cell array 31 is a block of cells that are simultaneously erased in a batch. Although not shown, the sources of the cell transistors are commonly connected in the array 31 and biased by the erase potential during erase. The common source is grounded during other operations such as writing and reading. On the other hand, the drains of the cell transistors are commonly connected to the bit lines arranged orthogonally to the word lines WL for each column. Since these drains are open as described above during erasing, no special decoding operation is required, so they are omitted here.

The match signal HIT output from the comparator 26 is also supplied to the spare row decoder 35. The spare row decoder 35 receives the match signal H.
A level shifter 36 that shifts IT to a level between the potential Vpr and the ground potential GND, and an operating power supply are composed of a buffer 37 including a CMOS inverter having the potential Vpr and the bias potential Vbb. The level shifter 36 and the buffer 37 have the same circuit configurations as the level shifter 33 and the buffer 34 shown in FIGS. 15 and 16, respectively. Then, the spare word line SWL is driven by the output of the buffer 37.

Next, the operation of the above structure will be schematically described. At the time of reading and writing, each word line WL in the memory cell array 31 has a row address RA.
It needs to be selected one by one according to dd. In the circuit shown in FIG. 14, a row address RAdd designated by an external input or an address counter inside the chip is decoded by the predecoder 25 and then further decoded by the main decoder 30 to select one word line WL. It is like this. The potential levels and word line WL levels are summarized in Table 1 below.

[0017]

[Table 1]

That is, the potential Vbb is at the ground level during reading and writing, only the word line selected by the row address signal RAdd is at the Vpr level, and the unselected word lines are at the potential Vbb (ground level). . on the other hand,
At the time of erasing, the activation signal PRE of the predecoder 25 becomes "H" level and the potential Vbb becomes a negative potential. Signal P
Since all the row predecode signals RPD are in the non-selected state by RE, all the word lines are fixed to the non-selected side,
In the output of the level shifter 33 (= input of the buffer 34), all the rows have the potential Vpr. Therefore, the P-channel MOS transistor Q in the buffer 34 in all rows is
5 is a non-conducting state, N-channel type MOS transistor Q6
Becomes conductive, and the word line WL is driven at the potential Vbb. Therefore, all the word lines WL are biased with a negative potential at the same time, and erase is performed collectively.

At this time, since the N-channel type MOS transistor Q6 in the buffer 34 needs to be driven at a negative potential, as shown in FIG. 17, when a P-type semiconductor substrate is used, this substrate 11 is used. Not in P-well region 40
It is provided inside. The P well region 40 is formed in the N well region 41 biased with the potential Vw and is separated from the substrate 11 at the ground level. The potential Vw is M
It is equal to or larger than the source potential Vs and the ground potential of the OS transistor Q6, that is, Vw ≧ V
The relationship of s and Vw ≧ GND is satisfied.

Now, consider the row redundancy in the above-mentioned flash memory. Since only the selected word line is at the “H” level during reading and writing, replacement with a spare cell does not select the corresponding word line when it is detected that the row address of the defective cell has been selected. This is performed by activating the spare word line SWL instead.

Next, the erase operation will be considered in detail.
Erasing is performed collectively on a block-by-block basis for cells whose sources are commonly connected. Therefore, even if the redundancy cell is used to replace the defective cell with the spare cell, the source is still physically connected, so that the erase potential is also applied to the source of the defective cell. In addition, FIG.
In the configuration as shown in FIG. 6, the potential Vbb becomes negative regardless of the row address during erasing, so that all word lines are biased with the negative erasing potential. That is, the defective cell replaced by redundancy and the spare row cell are applied with the same erase potential as the other cells to be erased. Unless the erase defective cell is erased, the threshold voltage Vth is lowered. Become. If the threshold voltage drops to a negative value and becomes over-erased, current flows even if the word line WL is unselected and is at the ground level, and the drain of this defective cell is connected to the bit line. The problem arises that it destroys the normal reading of other cells in the same column connected to.

In order to avoid this problem, all cells in the block including spares and defective rows are written before erasing, and the threshold voltage after erasing is made negative by making the threshold voltage high. It is being done to prevent it from happening. However, the defective cell is not guaranteed to be normally written, the threshold voltage before erasing cannot be increased, and the threshold voltage after erasing may become negative, which is a so-called overerase state. Therefore, according to the conventional erasing method, the redundancy circuit cannot be used to repair a defective cell.

Further, pre-erase programming for preventing over-erase must be performed on all cells including defective cells and spare rows. Now, it is assumed that all addresses are written while incrementing the addresses in order. At this time, if there is a defective row replaced by the redundancy circuit, this defective row is not selected by the address, and therefore writing is not performed. In addition, writing is not performed even on an unused spare row that has not been replaced. Therefore, in the pre-erase programming, in addition to writing to all addresses, special control is required to separately select a defective row and an unused spare row that are not normally selected for writing. However, the erase sequence is complicated. Further, writing using hot electrons cannot be performed in a batch for all cells, unlike erasing because of current limitation, and must be performed while incrementing the address in units of bytes or the like. Therefore,
The erasing time also increases as the number of writing cells increases (defective rows and spare rows).

Next, consider the case where a cell having a defective insulating film is replaced with a spare cell. Insulating film defects include interpoly insulating film leakage (floating gate 15 and control gate 1) as schematically shown by resistor R1 in FIG.
7) and a tunnel oxide film leak (floating gate 1) as schematically shown by the resistor R2 in FIG.
5 and a leak between the substrate 11). In any of the defects, electrons cannot be stored in the floating gate 15, but if the word line (control gate 17) is at the ground level, the drain current does not flow. The problem of read breakdown of other cell transistors due to such current leakage does not occur. Therefore, it seems that the spare cells can be relieved, and the normal operation can be performed by actually substituting the spare cells, resulting in a good product.

However, this kind of defect involves the most troublesome problem of destruction during repeated erasing and writing. As described above, the high potential for erasing is applied to the control gate 17 and the source 12 during erasing. At this time, the floating gate 15 is biased to a potential determined by the bias level of each terminal and the capacitance formed by the insulating films 14 and 16, and the amount of charges accumulated in the floating gate 15. That is, the potential difference between the source 12 and the control gate 17 is divided and applied to both insulating films 14 and 16.
However, in the case of the failure shown in FIG. 18, the negative potential of the control gate 17 is transmitted to the floating gate 15, and the high potential between the source 12 and the control gate 17 is directly applied to the tunnel oxide film 14. Therefore, by repeatedly erasing, breakdown of the insulating film (T
DDB) will happen. On the other hand, in the case of a defect as shown in FIG. 19, the same thing may occur on the interpoly insulating film 16 side. In either case, a leak path is formed between the control gate 17 and the substrate 11 via both insulating films 14 and 16.

In the case of erasing, all word lines are biased to a negative potential at the same time. However, since this negative potential is generated not by the outside of the chip but by the circuit inside the chip, the current supply capacity is limited. Therefore, if the negative potential is short-circuited with the substrate due to the leak between the control gate 17 and the substrate 11 in the defective row, there is a possibility that a sufficient negative potential cannot be obtained. Therefore, the negative potential applied to the word line of a normal row other than the defective row is also influenced, and the erasing failure of the normal cell in the same block occurs. Therefore, it is not possible to save the defects in bit units due to the insulating film defects by the spare row.

The above-mentioned defective writing, defective insulating film, and the like are major symptoms of defective cells, and if a redundancy circuit cannot be repaired, the repair rate is greatly reduced, and the redundancy circuit is adopted. It causes a problem that the effect is reduced. In addition, since it is necessary to perform the pre-erase writing on the defective row and the unused spare row, the erase sequence is complicated and the time is increased.

[0028]

As described above, the conventional non-volatile semiconductor memory device has a problem that even if the redundancy circuit is provided, it is not possible to sufficiently remedy a cell having a defective insulating film or a cell having a defective writing.

Further, since it is necessary to perform the pre-erase writing on the defective row and the unused spare row, there is a problem that the erase sequence is complicated and the time is increased. The present invention has been made in view of the above circumstances, and an object of the present invention is to provide a nonvolatile semiconductor memory device capable of reliably relieving a cell having a defective insulating film or a cell having a defective writing. It is in. Another object of the present invention is to
It is an object of the present invention to provide a non-volatile semiconductor memory device capable of simplifying the erase sequence and shortening the time.

[0030]

According to another aspect of the present invention, there is provided a nonvolatile semiconductor memory device having redundancy means for repairing a defective cell by replacing it with a spare cell, and a defect for storing an address of the defective cell to be replaced with the spare cell. It is characterized by comprising address storage means and transfer means for transferring the address of the defective cell stored in the defective address storage means to the address bus inside the chip at the time of erasing.

According to another aspect of the non-volatile semiconductor memory device of the present invention, redundancy means for replacing a defective cell with a spare cell for repair, defective address storage means for storing an address of a defective cell to be replaced with a spare cell, and latching an input address. The address latch means for transferring the latched address to the address bus inside the chip and the transfer means for transferring the address stored in the defective address storage means to the address latch means are provided. It is characterized in that the address of the defective cell latched by the address latch means is transferred to the address bus inside the chip.

According to another aspect of the non-volatile semiconductor memory device of the present invention, redundancy means for replacing a defective cell with a spare cell for repairing, defective address storage means for storing an address of a defective cell to be replaced with a spare cell, and row decode signals for each row. Selected by the address of the defective cell stored in the defective address storage means prior to erasing, the first latch means for latching the spare cell, the second latch means for latching a signal for selecting a spare cell for each spare row. Regarding the second latch means corresponding to the unused row of the first latch means and the spare row corresponding to the row, the use of the first latch means and the spare row corresponding to the non-defective row is performed. Means for setting the opposite state of selection / non-selection to the second latch means corresponding to the row Provided, and wherein the erasing on the basis of the data latched in the first, second latch means.

A non-volatile semiconductor memory device according to a fourth aspect of the present invention includes a redundancy circuit for repairing a defective cell by replacing it with a spare cell, and a flash memory for erasing data by applying a negative bias to a control gate of a cell transistor. At defective row address storage means for storing the row address of the defective cell, address comparison means for comparing the address output to the address bus inside the chip with the row address stored in the defective row address storage means, and storage data. At the time of erasing, when the address comparison means detects an address match, the spare row decoding means for selecting a spare row cell is controlled, and the defective row address output to the address bus inside the chip is transferred to the row decoding means. Control means and transfer by this transfer control means Characterized by comprising a voltage applying means for applying a ground potential to the word line of the defective row address based on the decoded signal of the row address by row decode means.

A nonvolatile semiconductor memory device according to a tenth aspect of the present invention is provided with a redundancy circuit for repairing defective cells by replacing them with spare cells, and a flash memory for erasing by applying a negative bias to the control gates of the cell transistors. , An address buffer to which an address signal is input, an address latch circuit that latches the address signal input to this address buffer, an address counter that generates an address signal, a redundancy ROM that stores a defective row address, and the above address. Outputs of the buffer, the address latch circuit, the address counter, and the redundancy ROM are supplied, and an address multiplexer that selectively outputs these outputs to an address bus inside the chip, and a row add output to the address bus. Graphics and a comparator for comparing the defective row address stored in the redundancy ROM,
A row predecoder for decoding the row address output to the address bus, a main decoder to which the row predecode signal output from the row predecoder is supplied, and a word line selected by the output of the main decoder Memory cell array in which a row of memory cells is selected by, a spare row cell for replacing a defective row in the memory cell array, a spare row decoder for selecting this spare row cell, and a row address match by the comparator when erasing stored data. When detected, the main row decoder activates the row predecoder and drives the spare row decoder to control the spare row cells. On the word line to which the cell is connected Comprising a first biasing means for applying a ground potential, said spare row decoder, the applying the ground potential to an unused spare word line during the erase of the stored data 2
It is characterized in that it is provided with a biasing means.

The non-volatile semiconductor memory device according to claim 11 is
In a flash memory that includes a redundancy circuit that replaces a defective cell by repairing it by a spare cell, and applies a negative bias to the control gate of a cell transistor to erase, an address buffer to which an address signal is input and an address counter that generates the address signal. An address latch circuit for selectively latching the address signal input to the address buffer and the address signal generated by the address counter, a redundancy ROM for storing a defective row address, the address buffer,
Outputs of the address latch circuit and the address counter are supplied, and an address multiplexer that selectively outputs these outputs to an address bus inside the chip, a row address output to the address bus, and the redundancy ROM are stored. Compare with defective row address,
When a match occurs, a match signal is output to control the address latch circuit, and a comparator for latching the row address generated by the address counter in the address latch circuit and a row predecode for decoding the row address output to the address bus A decoder, a main decoder to which a row predecode signal output from the row predecoder is supplied, a memory cell array in which a row of memory cells is selected by selecting a word line by the output of the main decoder, A spare row cell for replacing a defective row in the memory cell array, a spare row decoder for selecting this spare row cell, and a row predecoder are activated and the spare predecoder is activated when a match is detected by the comparator when erasing stored data. Drive row decoder And a logic circuit for controlling a spare row cell, the main row decoder including first bias means for applying a ground potential to a word line connected to a defective cell in the memory cell array when erasing stored data. The row decoder is characterized by including second bias means for applying a ground potential to an unused spare word line when erasing stored data.

According to a thirteenth aspect of the non-volatile semiconductor memory device,
In a flash memory that includes a redundancy circuit that replaces a defective cell with a spare cell and repairs it by applying a negative bias to the control gate of a cell transistor, an address buffer to which an address signal is input, and an address buffer to which the address signal is input. An address latch circuit that latches an address signal, an address counter that generates an address signal, a redundancy ROM that stores a defective row address, an output of the address buffer, the address latch circuit, and the address counter are supplied, and the selected address is supplied. Address multiplexer for outputting the address to the address bus inside the chip, the row address output to the address bus, and the redundancy ROM
A comparator for comparing with the defective row address stored in, a row predecoder for decoding the row address output to the address bus, a main decoder for decoding the row predecode signal output from the row predecoder, A first latch circuit that latches the decode signal of the main decoder for each row, and a word line is selected based on the decode signal latched by the first latch circuit to select the row of the memory cell. A memory cell array; a spare row cell for replacing a defective row of the memory cell array; a spare row decoder for selecting the spare row cell; and a second latch circuit for latching a decode signal output from the spare row decoder for each spare row. , When deleting stored data, the above comparator And a logic circuit for activating the row predecoder and setting a second latch circuit in the spare row decoder to control a spare row cell when an address match is detected. Includes first bias means for applying a ground potential to a word line to which a defective cell in the memory cell array is connected at the time of erasing stored data, and the spare row decoder supplies an unused spare word line at the time of erasing the stored data. It is characterized in that it is provided with a second bias means for applying a ground potential.

[0037]

According to the first, second, and third configurations, when erasing is performed by applying a negative potential to the control gate of the cell transistor during erasing (SGE method),
By transferring the address of the defective cell stored in the defective address storage means to the address bus, the defective row can be fixed in the selected state, so that the application of a negative potential to the control gate of the cell transistor of the defective row can be avoided. Therefore, it is possible to surely relieve a cell having a defective insulating film or a cell having a defective writing. By providing the first and second latch means as described in claim 3, even when there are a plurality of defective rows, it is possible to avoid applying a negative potential to the control gates of the cell transistors of the plurality of defective rows. .

According to the fourth, tenth, eleventh and thirteenth aspects, the potential applying means or the first bias circuit is provided to the defective row word line, that is, the control gate of the defective cell transistor. Since erasing is performed in the state where the ground potential is applied from the above, it is possible to reliably rescue cells with defective insulating films and defective writing. Further, as in the tenth, eleventh and thirteenth aspects, if the ground potential is applied to the unused spare word line from the second bias circuit, the pre-erase writing need not be performed.
The erase sequence can be simplified and the time can be shortened. Furthermore,
When the first and second latch circuits are provided as described in claim 13, even when there are a plurality of defective rows, the problem of application of a negative potential to the control gates of the cell transistors of these defective rows and the erasing are eliminated. It is also possible to avoid the problem that the erase sequence is complicated and the time is increased by the pre-writing.

[0039]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS An embodiment of the present invention will be described below with reference to the drawings. FIG. 1 is a diagram for explaining a nonvolatile semiconductor memory device according to a first embodiment of the present invention, in which a row decode circuit of a flash memory and a peripheral circuit section related to an erase operation are extracted and shown. The address signal Add supplied from the outside is input to the address buffer 21, and the output of the address buffer 21 is supplied to the address multiplexer 22 and the address latch circuit 23. The output of the address latch circuit 23, the output of the address counter 24, and the data of the defective row address stored in the redundancy ROM 29 are respectively supplied to the address multiplexer 22, and are supplied to the address buffer 21, the address latch circuit 23, and the address counter 24. Internal address bus A when either output is selected
Row predecoder 25 and comparator 26 via B
Is supplied to. The row predecoder 25 is composed of AND gates 27, 27, ... And the row address signal RAdd of the address signals supplied via the internal address bus AB is supplied to each of them. Also,
An activation signal PRE of the row predecoder 25 output from the inverter 51 is supplied to each AND gate 27, 27, ...

The comparator 26 compares the address supplied via the internal address bus AB with the defective address stored in the redundancy ROM 29, and outputs a match signal HIT when they match. Redundancy RO
M29 is a defective address storage circuit for storing the address of a defective cell, and this ROM 29 can store as many bits as the number of row addresses in the case of row redundancy. On the other hand, if each word line is replaced, all row addresses are stored in the redundancy ROM 29. Further, if the n-th power of 2 is collectively replaced, such as 2 rows or 4 rows, the address to be stored is reduced by n bits. For this storage, writing and erasing of data may be performed using a flash memory cell, or a defective address may be stored by providing a fuse using polysilicon and fusing with a laser. Alternatively, instead of the fuse, an EPROM cell may be provided as shown in FIG. The defective address is stored in the redundancy ROM 29, and the comparator 26 constantly checks whether the selected address matches the defective address.
If the selected address matches the bad address,
The coincidence signal HIT becomes "H" level.

The match signal HIT is sent to the NAND gate 52.
Is supplied to one of the input terminals. An inverted signal of the erase signal ERS is supplied to the other input end of the NAND gate 52, and the output end is connected to one input end of the NAND gate 53. The erase signal ERS and the spare row activation signal S
The inverted signal of PEi is supplied to the input terminal of the NAND gate 54. The output of the NAND gate 54 is supplied to the other input terminal of the NAND gate 53. The output of the NAND gate 53 is supplied to the input terminal of the inverter 51 and the spare row decoder 55.

When the activation signal PRE output from the inverter 51 goes to "L" level, the predecoder 25 is deactivated to bring the defective row into a non-selected state, and the spare row decoder 55 spare word in the spare row cell 38. The line SWL is driven and replacement with a spare cell is performed. At the time of erasing, the word lines WL in the memory cell array 31 are collectively operated, all are in the non-selected state, and the spare rows are collectively erased. Therefore, control by redundancy is not particularly performed.

The row predecode signal RPD output from the row predecoder 25 is supplied to the main decoder 56. The main decoder 56 includes an AND gate 32, which corresponds to each word line WL of the memory cell array 31,
32, ..., First level shifters 57, .., Second level shifters 58 ,. The first level shifter 57 operates with the operating power supplies Vcc and Vbb, shifts the output signal of the AND gate 32 to these signal levels, and outputs it. The second level shifter 58 is
This is a circuit for converting a signal level in order to bring the word line WL to a high potential at the time of writing, converting a Vcc system signal into a Vpr system signal and outputting it.

The memory cell array 31 is a block of cells that are simultaneously erased collectively. Although not shown, the source of each cell transistor is commonly connected in the array 31, and an erase potential is applied during erase. The common source is grounded during other operations such as writing and reading. On the other hand, the drains of the cell transistors are commonly connected to the bit lines arranged orthogonally to the word lines WL, ... For each column. Since these drains are open during erasing as described above, no special decoding operation is necessary, and therefore they are omitted here.

Further, the spare row decoder 55 has each spare word line SWL (only one spare word line is representatively shown in FIG. 1 for the sake of simplicity).
When a plurality of spare word lines are used, decoding is performed using an AND gate, a NAND gate, etc.), a first level shifter 59, a second level shifter 60, and a buffer 37 are provided. First level shifter 59
Shifts the output signal of the NAND gate 53 to a level between the power supply voltage Vcc and the bias potential Vbb. The second level shifter 60 shifts the output signal of the first level shifter 59 to a level between the potential Vpr and the bias potential Vbb. The buffer 37 has a CMOS inverter whose operating power source has a potential Vpr and a bias potential Vbb.
At 7, the spare word line SWL in the spare row cell 38 is driven. The level shifters 59 and 60 and the buffer 3
7 has substantially the same circuit configuration as the level shifters 57 and 58 and the buffer 34 in the main row decoder 56.

FIG. 2 shows a configuration example of the redundancy ROM 29 in the circuit shown in FIG. The redundancy ROM 29 includes an EPROM cell 70, a latch circuit 71, a selector circuit 72, and capacitors C1 and C2. The latch circuit 71 is a CMO.
It is composed of an S inverter circuit IV1 and a P channel MOS transistor T1 for feedback. The source of the transistor T1 is connected to the power supply Vcc, the drain is connected to the input node of the inverter circuit IV1, and the gate is connected to the output node of the inverter circuit IV1.
A capacitor C1 is connected between the input node of the inverter circuit IV1 and the power source Vcc, and the capacitor C1 is connected to the inverter circuit IV1.
A capacitor C2 is connected between the output node of 1 and the ground point Vss. The selector circuit 72 includes a CMOS inverter circuit IV2 and two CMOS transfer gates CT.
1 and CT2. This selector circuit 72 is
The 1-bit address signal Adi or its inverted signal / Adi is selected and output according to the latch data of the latch circuit 71. The control gate of the EPROM cell 70 has a high voltage Vpp at the time of writing for redundancy.
The ground potential Vss is applied during normal operation.

FIG. 3 shows the first circuit in the circuit shown in FIG.
3 shows an example of the configuration of the level shifters 57 and 59 of FIG. Here, the structure of the level shifter 57 will be described as an example.
The level shifter 59 has the same structure. Level shifter 57
Is composed of P-channel MOS transistors Q7 and Q8, N-channel MOS transistors Q9 and Q10, and an inverter 73. MOS transistor Q7,
The source of Q8 is connected to the power source Vcc, and between each drain and the bias potential Vbb, MOS transistors Q9,
The drain and source of Q10 are connected. The gate of the MOS transistor Q9 has the MOS transistor Q8,
The drain of Q10 is connected to the common connection point, and the gate of MOS transistor Q10 is
It is connected to the common drain connection point of Q9. The output signal of the AND gate 32 is supplied to the gate of the MOS transistor Q7, and the output signal of the AND gate 32 is supplied to the gate of the MOS transistor Q8 via the inverter 73. Then, the level shifter 58 outputs the output signal obtained from the connection point of the MOS transistors Q8 and Q10.
It is designed to be supplied to.

FIG. 4 shows a second circuit in the circuit shown in FIG.
3 shows an example of the configuration of the level shifters 58 and 60 of FIG. Here, the level shifter 58 will be described as an example. The level shifter 33 includes a P-channel type MOS transistor Q11,
Q12, N-channel type MOS transistors Q13, Q1
4 and an inverter 74. The sources of the MOS transistors Q11 and Q12 are respectively connected to the potential Vpr, and a MOS is connected between each drain and the bias potential Vbb.
The drains and sources of the transistors Q13 and Q14 are connected. The gate of the MOS transistor Q11 is the above M
It is connected to the drain common connection point of the OS transistors Q12 and Q14, and the gate of the MOS transistor Q12 is connected to the drain common connection point of the MOS transistors Q11 and Q13. The output signal of the level shifter 57 is supplied to the gate of the MOS transistor Q13,
A level shifter 57 is provided at the gate of the OS transistor Q14.
Output signal is supplied via the inverter 74. The output signal obtained from the connection point of the MOS transistors Q12 and Q14 is supplied to the buffer 34.

FIG. 5 shows a configuration example of the buffers 34 and 37 in the circuit shown in FIG. Here, the buffer 34 will be described as an example. The buffer 34 is P
Channel type MOS transistor Q15 and N channel type M
It is composed of a CMOS inverter including an OS transistor Q16. The operating power supply of this CMOS inverter is the potential Vpr2 and the bias potential Vbb, and the output of this buffer 34 drives the corresponding word line WL in the memory cell array 31. Table 2 below shows each potential level and word line level in the above-described circuit.

[0050]

[Table 2]

In the first embodiment shown in FIGS. 1 to 5, in the case of reading and writing, the same control as that of the circuit shown in FIG. 14 may be performed. In other words, when a row containing a defective cell is selected, this defective row is in the non-selected state,
That is, the word line of the defective cell is not biased because it is fixed to the ground level.

Next, the operation at the time of erasing will be described. At the time of erasing, the output of the redundancy ROM 29 is selected by the address multiplexer 22, and the data of the defective row address stored in this ROM 29 is directly transferred to the address bus AB. In the case of a flash memory capable of automatic operation under command control, as shown in FIG. 14, in order to transfer the external input address Add, the output of the address latch circuit 23, the output of the address counter 24, etc. to the internal address bus AB by switching. Is provided with an address multiplexer 22. Therefore, in the first embodiment, the data stored in the redundancy ROM 29 is input to the multiplexer 22.

The row predecoder 25 is shown in FIG.
Although it is inactivated in the circuit shown in FIG. 4, in the present invention, decoding is performed in accordance with the signal of the address bus AB as usual. That is, the address of the defective cell is decoded. The first level shifter 57 in the main decoder 56 is a circuit for shifting the "L" level side of the Vcc system signals ("H" level = Vcc, "L" level = ground potential) to Vbb level. The second level shifter 58 is "H".
This is a circuit for shifting the potential on the level side from Vcc to Vpr, and the bias potential Vbb is on the "L" level side.
These two-stage level shifters 57 and 58 change the Vcc system signal output from the AND gate 32 to Vpr / Vcc.
It is converted into a system signal and input to the buffer 34 (however, as shown in Table 2, the potential Vpr at the time of erasing is at the same level as the power source potential Vcc).

The power supply of the buffer 34 for driving the word line WL is the potentials Vpr2 and Vbb. Vpr2 is different from Vpr at the ground level at the time of erasing, but operates as an inverter according to the output signal of the level shifter 58. That is, when the output of the level shifter 58 is the potential Vpr, the Vbb level is output, and when the output of the level shifter 58 is the potential Vbb, Vpr.
(= Ground level) is output. The row designated by the internal address bus AB is selected, and the word line WL is Vpr.
= Ground level. Here, since the defective row address is selected, only the defective row becomes the ground level and all the other rows become the Vbb level which is the erase potential. Therefore,
The stress application to the control gate of the defective cell is eliminated, and the above-mentioned problem of the insulating film breakdown over time can be avoided.
On the other hand, with respect to the spare row, the negative potential Vbb is applied to the word line WL at the time of erasing and the unused spare row is fixed to the ground level only when the redundancy replacement is performed.

The above control is performed by the erase signal ER shown in FIG.
S, the spare row activation signal SPEi, and the match signal HIT between the selected address and the defective address. By the way, the sources of the cell transistors in the memory cell array 31 are all common in the block. Therefore, when a defective cell exists, the erase potential is also applied to the source of the defective cell during erasing even after replacement with the spare cell. However, in the case of the SEG method, the source potential is considerably lower than that in the SE method (for example, about 5 V in the SGE method compared to about 12 V in the SE method), and in order to maintain an electric field sufficient to pass the tunnel current, It is necessary to apply a negative potential to the gate. Therefore, even if the source of the defective cell is biased, it is not erased like other cells. On the contrary, it is known that when the source is biased with the gate and drain grounded, the amount of charge in the floating gate converges to a stable state. Therefore, by properly matching the coupling of the insulating film and the bias level, the defective cell can be expected to converge to a certain threshold voltage having a stable positive value, not to decrease the threshold voltage due to erasing. Therefore, if all the bit lines are grounded and the drains are grounded, the control gates of the defective row and the unused spare row are grounded, so even if the erase potential is applied to the source during erase,
The threshold voltage settles to some positive value. Therefore, the defective cell and the spare row are not erased, and pre-erase writing to these cells becomes unnecessary. As a result, even a defective cell can be repaired by the spare row. Further, the erase sequence can be simplified and the time can be shortened.

FIG. 6 is for explaining a nonvolatile semiconductor memory device according to the second embodiment of the present invention.
The row decode circuit in the flash memory and the peripheral circuit section related to the erase operation are extracted and shown. 6, the same components as those in FIG. 1 are designated by the same reference numerals and detailed description thereof will be omitted. That is, in the second embodiment, the coincidence signal HIT output from the comparator 26 is supplied to the address latch circuit 23. The address latch circuit 23 includes an address counter 24
The address signal output from the address buffer 21 and the address signal output from the address buffer 21 are supplied as in the first embodiment, and the coincidence signal HIT controls the data fetch operation of the address latch circuit 23.

The level of each potential and the level of the word line are shown in Table 2.
Is the same as. Similar to the first embodiment, the defective row and the unused spare row are fixed to the ground level by setting the internal address to the selected state of the defective address at the time of erasing. Therefore, the decoder is exactly the same. The difference from the first embodiment is the method of transferring the defective address to the address bus AB.

The control in this embodiment will be described in order with reference to the timing chart of FIG. When the count-up signal is supplied to the address counter 24 before erasing, the row address is sequentially incremented by the address counter 24 and the address signal is supplied to the address bus AB via the address multiplexer 22. Row address output to address bus AB and redundancy RO
The comparator 26 checks the match with the defective row address stored in M29.
Outputs a match signal HIT of "H" level. When the HIT signal goes to "H" level, the address output from the address counter 24 becomes the address latch circuit 23.
Latched on. Then, when the erase signal ERS becomes "H" level, the address latched in the address latch circuit 23 is selected by the address multiplexer 22 and transferred to the address bus AB, and the address bus AB is brought into the selected state of the defective address. As a result, the potential of the defective word line is fixed to the ground GND level, and the potential of the spare word line becomes a negative potential. On the other hand, when a normal word line is selected, a negative potential is applied.

According to the above-mentioned structure, compared with the conventional circuit, it is possible to relieve a cell having a defective writing or a cell having a defective insulation film destruction without providing an additional circuit, and a defective product using the redundancy technique can be provided. The relief rate can be greatly improved. A sequence of defective address check by incrementing the address is required, but since the sequence is added, the circuit scale hardly increases. Further, in the first embodiment, since the data stored in the redundancy ROM 29 is directly input to the address multiplexer 22, it is necessary to provide signal lines for the number of row addresses between both circuits. Therefore, it is 1 for megabit class memory.
When the number of circuits is close to 0 and both circuits are separated, a large wiring area may be required. In such a case, the circuit of the circuit shown in the second embodiment is more important than that of the first embodiment. Is more advantageous.

FIG. 8 is for explaining a nonvolatile semiconductor memory device according to the third embodiment of the present invention.
The row decode circuit in the flash memory and the peripheral circuit section related to the erase operation are extracted and shown. This Figure 8
In the circuit shown in FIG. 1, latch circuits (R / S flip-flops) 61, ..., 61 are provided at the input ends of the level shifters 57, ... In the circuit shown in FIG. S flip-flop) 62 is provided. The flip-flop 61,
..., AND gate 63 at the set input terminal S of 61 ,.
The decode signal output from 63 is supplied, and the inverted signal / LAH of the latch signal LAH is supplied to the reset input terminal R. Then, these flip-flops 61, ...
61 outputs / Q are supplied to the input terminals of the level shifters 57 ,. A row predecode signal RPD and a reset signal RST are supplied to the input ends of the AND gates 63, ..., 63.

The set input terminal S of the flip-flop 62 is supplied with the output signal of the OR gate 64, and the reset input terminal R is supplied with the output signal of the AND gate 65. The output / Q of the flip-flop 62 is supplied to the input terminal of the level shifter 59. Or gate 64
The reset signal RST is supplied to one input end, and the output signal of the AND gate 66 is supplied to the other input end. The latch signal LAH is supplied to one input end of the AND gate 65, and the spare row activation signal SPEi is supplied to the other input end. A row predecoder activation signal PRE output from the exclusive NOR gate 67 is supplied to the first input terminal of the AND gate 66, and the spare row activation signal SP is supplied to the second input terminal.
Ei is supplied, and the latch signal LA is applied to the third input terminal.
An inverted signal of H is supplied. Further, the comparator 26 is connected to one input terminal of the exclusive NOR gate 67.
The coincidence signal HIT output from the above is supplied, and the latch signal LAH is supplied to the other input terminal. Since the latch signal LAH is the inversion of the exclusive OR of the coincidence signal HIT, the activation signal PRE of the row predecoder 25 is "H" when both signals coincide.
The level becomes "L" level when they do not match.

The level of each potential and the level of the word line are shown in Table 2.
Is the same as. The defective row and the unused spare cell are the same as those in the first embodiment in that the gate is fixed to the ground potential at the time of erasing. In this embodiment, the main decoder 56 is provided with latch circuits 61, ..., 61 for each word line WL so that selection and non-selection can be latched for each word line WL. Therefore, it is only necessary to set the latch circuit 61 to the selected state only for the row having the defective cell which is not desired to be erased and the unused spare row, and to set the other rows to be erased to the non-selected state before starting the erase operation.

The operation of the circuit shown in FIG. 8 will be described below. First, consider modes other than erase. In the modes other than the erase mode, the latch signal LAH and the reset signal RST
Is fixed at "L" level. In the latch circuit 61 in the main decoder 56, the latch signal / L which is a reset input is used.
Since AH is fixed at the "H" level, an inverted signal of the decode input is output as the output / Q, and the latch circuits 61, ...
1 simply performs an inverter operation. Therefore, when the decode signal becomes "H" level, the word line WL is in the selected state. Next, consider the case where the coincidence signal HIT is at "L" level. At this time, the activation signal PRE becomes "H" level, so that the predecoder 25 makes a selection according to the address. Therefore, the word line WL specified by the address signal
Is selected. At this time, the spare row is in a non-selected state.

When the match signal HIT is at "H" level, that is, when a defective cell is selected, the activation signal PRE of the predecoder 25 becomes "L" level, the predecoder 25 is deactivated and the selected word line is It will be unselected. on the other hand,
Since the spare row activation signal SPEi is at the "H" level, the designated spare row to be replaced is selected.
Therefore, a desired operation is performed in which the spare row is replaced only when the defective address is selected.

Next, the erase mode will be described with reference to the timing chart of FIG. When erasing,
First, the reset signal RST is set to "H" level. As a result, all the word lines WL are in the non-selected state and all the spare rows are in the selected state. Latch signal LA in this state
When H is set to "H" level, the latch mode is entered. The latch signal LAH goes to "H" level, so that the flip-flops 61, ..., 61 of each word line start the latching operation, but at this time, each word line is in the non-selected state and all the spare word lines are in the selected state. Latched. After the latch is completed, the reset signal RST is returned to the “L” level. Next, similarly to the second embodiment, the address counter 24 sequentially increments the row address, and the comparator 26 checks whether or not the row address matches the defective row address stored in the redundancy ROM 26. When they do not match, the match signal HIT is at the “L” level, so the activation signal PIT
RE becomes "L" level, the predecoder 25 is inactivated, and all word lines (excluding spare word lines) WL
Is in a non-selected state. Therefore, the latch data of the word line does not change. Match signal HIT is "H" only when they match
However, at this time, the activation signal PRE becomes "H" level and the predecoder 25 is activated. Therefore, the decode signal of the word line selected by the address at this time, that is, the defective row address becomes "H" level, and the latch data of the word line is changed to the selected state and held. FIG. 9 shows a case where there are two defective row addresses, and defective address 1 and defective address 2 are present.
When the coincidence signal HIT is output in correspondence with, the / Q of the defective rows 1 and 2 corresponding to these addresses are both "L".
The normally low / Q output is at the "H" level. On the other hand, for the spare row, the defective address is selected,
When the activation signal SPEi of the spare row to be replaced becomes "H" level, the latch circuit 62 changes from the selected state to the non-selected state and is held. That is, the / Q output of the spare row corresponding to the defective address 1 becomes "H" level as shown by the solid line, and the / Q output of the spare row corresponding to the defective address 2 becomes "H" level as shown by the broken line,
The / Q output of an unused spare row is fixed at "L" level.

Therefore, after incrementing all the addresses, the latch circuits of the defective row and the unused spare row are held in the selected state (latch state), and the other rows are held in the non-selected state (reset state). If erasing is performed while holding the latched data in this way, the output / Q of the latch circuits 61 and 62 is output during the period when the erase signal is at the "H" level.
The erase potential is applied to the word line and the spare word line whose level is "H", and the ground potential is applied to the word line and the spare word line whose level / Q is "L".

In the third embodiment described above, it is necessary to provide a latch circuit for each word line. However, even if a plurality of defects are randomly present in the same block, erasure of the word line and spare word line can be performed. Therefore, the effect of avoiding the application of a negative potential can be obtained. The address increment in the latch mode may be performed in the same manner as in the second embodiment.

In each of the embodiments, the case where the word lines are controlled in units of one line has been described as an example. However, when replacing the spare row in units of a plurality of lines, the selection or selection in units of a plurality of spare lines is performed. It suffices to control non-selection.

[0069]

As described above, according to the present invention, it is possible to relieve a defective writing cell or a defective insulation film destruction cell which cannot be relieved even by using a spare row in a conventional flash memory, and the redundancy is provided. The repair rate of defective products using technology can be greatly improved. In addition, the erase sequence can be simplified and the time can be shortened.

[Brief description of drawings]

FIG. 1 is a circuit diagram for illustrating a nonvolatile semiconductor memory device according to a first embodiment of the present invention, in which a row decode circuit of a flash memory and a peripheral circuit section related to an erase operation are extracted and shown.

FIG. 2 is a redundancy ROM in the circuit shown in FIG.
3 is a circuit diagram for explaining an example of the configuration of FIG.

3 is a circuit diagram showing a configuration example of a first level shifter in the circuit shown in FIG.

4 is a circuit diagram showing a configuration example of a second level shifter in the circuit shown in FIG.

5 is a circuit diagram showing a configuration example of a buffer in the circuit shown in FIG.

FIG. 6 is a circuit diagram for illustrating a nonvolatile semiconductor memory device according to a second embodiment of the present invention, in which a row decode circuit of a flash memory and a peripheral circuit section related to an erase operation are extracted and shown.

7 is a timing chart for explaining the operation of the circuit shown in FIG.

FIG. 8 is a circuit diagram for illustrating a nonvolatile semiconductor memory device according to a third embodiment of the present invention, in which a row decode circuit of a flash memory and a peripheral circuit section related to an erase operation are extracted and shown.

9 is a timing chart for explaining the operation of the circuit shown in FIG.

FIG. 10 is a cross-sectional view showing a memory cell of a flash memory.

FIG. 11 is a sectional view for explaining a general erasing method of a flash memory.

12 is a sectional view showing a configuration example of a memory cell when the erasing method shown in FIG. 11 is adopted.

FIG. 13 is a sectional view for explaining another erasing method in the flash memory.

FIG. 14 is a circuit diagram for explaining a conventional nonvolatile semiconductor memory device, in which a row decode circuit in a flash memory and a peripheral circuit portion related to an erase operation are extracted and shown.

15 is a circuit diagram showing a configuration example of a level shifter in the circuit shown in FIG.

16 is a circuit diagram showing a configuration example of a buffer in the circuit shown in FIG.

17 is an N-channel type M in the circuit shown in FIG.
3A and 3B are cross-sectional views each illustrating a structural example of an OS transistor.

18 is a cross-sectional view illustrating an example of a defective insulating film of a cell transistor in the circuit illustrated in FIG.

FIG. 19 is a cross-sectional view for explaining another example of a defective insulating film of a cell transistor in the circuit shown in FIG.

[Explanation of symbols]

21 ... Address buffer, 22 ... Address multiplexer, 23 ... Address latch circuit, 24 ... Address counter, 25 ... Row predecoder, 26 ... Comparator, 2
9 ... Redundancy ROM, 31 ... Memory cell array, 3
4, 37 ... Buffer, 38 ... Spare row cell, 55 ... Spare row decoder, 56 ... Main row decoder, 57,
57 ... 1st level shifter, 58, 60 ... 2nd level shifter, 61, 62 ... R / S flip-flop, Add ...
Address signal, RAdd ... Row address signal, RPD ... Row predecode signal, PRE ... Predecoder activation signal, HIT ... Match signal, ERS ... Erase signal, SPEi ...
Spare row activation signal, WL ... Word line, SWL ... Spare word line.

─────────────────────────────────────────────────── ─── Continuation of the front page (51) Int.Cl. ⁶ Identification number Internal reference number FI Technical indication H01L 27/115 21/8247 29/788 29/792 H01L 27/10 434 29/78 371

Claims (14)

[Claims] 1. Redundancy means for replacing a defective cell with a spare cell for repair, defective address storage means for storing an address of the defective cell to be replaced with a spare cell, and address of the defective cell stored in the defective address storage means at the time of erasing. A non-volatile semiconductor memory device comprising: a transfer unit that transfers the data to an address bus inside the chip. 2. Redundancy means for replacing a defective cell with a spare cell for repair, defective address storage means for storing an address of a defective cell to be replaced with a spare cell, latching an input address, and latching the latched address inside the chip. Address latch means for transferring to the address bus of the defective cell, and transfer means for transferring the address stored in the defective address storage means to the address latch means, and the defective cells latched in the address latch means at the time of erasing data A nonvolatile semiconductor memory device, which transfers an address to an address bus inside a chip. 3. Redundancy means for repairing defective cells by replacing them with spare cells, defective address storage means for storing addresses of defective cells to be replaced with spare cells, and first latch means for latching row decode signals row by row. Second latch means for latching a signal for selecting a spare cell for each spare row, and the first latch corresponding to the row selected by the address of the defective cell stored in the defective address storage means prior to erasing Regarding the second latch means corresponding to the unused row of the means and the spare row, the second latch means corresponding to the used row of the first latch means and the spare row corresponding to the non-defective row. Selectable for Latch Means /
A non-volatile semiconductor memory device comprising means for setting a reverse state of non-selection, and erasing is performed based on the data latched by the first and second latch means. 4. A defective row address memory for storing a row address of a defective cell in a flash memory including a redundancy circuit for replacing a defective cell with a spare cell for repairing, and applying a negative bias to a control gate of a cell transistor for erasing. Means, an address comparing means for comparing the address output to the address bus inside the chip with the row address stored in the defective row address storing means, and the address comparing means detecting the coincidence of the address when the stored data is erased. Transfer control means for controlling the spare row decoding means for selecting a spare row cell and transferring the defective row address output to the address bus inside the chip to the row decoding means, and the row address transferred by this transfer control means. Signal decoded by row decoding means And a potential applying means for applying a ground potential to the word line of the defective row address based on the above. 5. When erasing, when address matching is detected by the address comparing means, there is provided address selecting means for selecting a defective row address stored in the defective address storing means and transferring it to the address bus. The non-volatile semiconductor memory device according to claim 4. 6. When erasing, when address matching is detected by the address comparing means, address latching means for latching the address generated by the address generating means, and the address latched by the address latching means are used as the address. 5. The non-volatile semiconductor memory device according to claim 4, further comprising an address selecting unit that transfers the data to a bus. 7. The potential applying means comprises first level shift means for shifting the low level side of the decode signal to a first potential, and second level shift means for the high level side of the output signal of the first level shift means. Second level shift means for shifting the potential of the second level shift means and an output signal of the second level shift means are supplied, and the second level shift means operates with a voltage between the third potential and the first potential to drive the word line. 7. The non-volatile semiconductor memory device according to claim 4, further comprising one buffer means. 8. A third level shift means for shifting the low level side of the output signal of the spare row decoding means to the first potential, and a high level side of the output signal of the third level shift means. The fourth level shift means for shifting to the second potential and the output signal of the fourth level shift means are supplied and operate at a voltage between the third potential and the first potential to operate the word line. The non-volatile semiconductor memory device according to claim 7, further comprising a second buffer unit that is driven. 9. A first latch circuit for latching a row decode signal, and a second latch circuit for latching a spare row decode signal.
Latch means of the first row, the first latch means corresponding to the row selected by the address stored in the defective row address storage means prior to erasing, and the second row corresponding to the unused row of the spare row. 9. The non-volatile semiconductor memory according to claim 4, wherein the latch means is set to a latched state, and erasing is performed according to the data latched by the first and second latch means. apparatus. 10. A flash memory comprising a redundancy circuit for replacing a defective cell with a spare cell for repairing, wherein a negative bias is applied to a control gate of a cell transistor for erasing, an address buffer to which an address signal is inputted, and an address buffer An address latch circuit for latching an address signal input to a buffer, an address counter for generating an address signal, a redundancy ROM for storing a defective row address, the address buffer, the address latch circuit, the address counter and the redundancy ROM. Address multiplexer for supplying these outputs to the address bus inside the chip, a row address output to the address bus, and a defective row address stored in the redundancy ROM. Comparator, a row predecoder for decoding the row address output to the address bus, a main decoder to which a row predecode signal output from the row predecoder is supplied, and an output of the main decoder. A memory cell array in which a row of memory cells is selected by selecting a word line in, a spare row cell for replacing a defective row in the memory cell array, a spare row decoder for selecting the spare row cell, and a stored row erase When a row address match is detected by the comparator,
A logic circuit for activating the row predecoder and controlling the spare row cell by driving the spare row decoder is provided. The main row decoder is connected to a defective cell in a memory cell array when erasing stored data. The spare row decoder is provided with a first bias means for applying a ground potential to the word line, and the spare row decoder is provided with a second bias means for applying a ground potential to an unused spare word line at the time of erasing stored data. Semiconductor memory device. 11. A flash memory, comprising a redundancy circuit for replacing a defective cell with a spare cell for repairing, wherein a negative bias is applied to a control gate of a cell transistor for erasing, an address buffer to which an address signal is input, and an address signal. Generating an address counter, an address latch circuit for selectively latching the address signal input to the address buffer and the address signal generated by the address counter, a redundancy ROM for storing a defective row address, and the address buffer. An address multiplexer supplied with the outputs of the address latch circuit and the address counter and selectively outputting these outputs to an address bus inside the chip; a row address output to the address bus; and the redundancy. A comparator that compares the defective row address stored in the ROM, outputs a coincidence signal when coincident, controls the address latch circuit, and causes the address latch circuit to latch the row address generated by the address counter; A row predecoder that decodes a row address that is output to the address bus, a main decoder that is supplied with a row predecode signal that is output from this row predecoder, and a word line is selected by the output of this main decoder. A memory cell array in which a row of memory cells is selected, a spare row cell for replacing a defective row in the memory cell array, a spare row decoder for selecting the spare row cell, and when a match is detected by the comparator when erasing stored data , Use the row predecoder And a logic circuit that drives the spare row decoder to control the spare row cells, and the main row decoder applies a ground potential to a word line connected to a defective cell in the memory cell array at the time of erasing stored data. A non-volatile semiconductor memory device comprising: a first bias means for applying the bias voltage; and the spare row decoder includes a second bias means for applying a ground potential to an unused spare word line when erasing stored data. 12. The first bias means shifts a low level side of a decode signal to a bias potential, and a high level side of an output signal of the first level shifter to a first potential. The second level shifter for operating the word line and the second level shifter for supplying the output signal of the second level shifter, operating at a voltage between the second potential and the bias potential, and driving the word line. The bias means includes a third level shifter that shifts the low level side of the decode signal to the bias potential, and a fourth level shifter that shifts the high level side of the output signal of the third level shifter to the first potential. A second buffer that operates at a voltage between the second potential and the bias potential and drives a spare word line, wherein the bias potential is
It is a ground potential during reading and writing, a negative potential during erasing, the first potential is a power source potential during reading and erasing, and a high potential for writing during writing, and the second potential.
12. The non-volatile semiconductor memory device according to claim 10, wherein the potential is a power supply potential during reading, a high potential for writing during writing, and a ground potential during erasing. 13. A flash memory comprising a redundancy circuit for replacing a defective cell with a spare cell for repairing, wherein a negative bias is applied to a control gate of a cell transistor for erasing, an address buffer to which an address signal is input, and an address buffer An address latch circuit for latching the address signal input to the buffer, an address counter for generating the address signal, a redundancy ROM for storing a defective row address, an output of the address buffer, the address latch circuit and the address counter are supplied. An address multiplexer for outputting the selected address to the address bus inside the chip,
A comparator that compares the row address output to the address bus with a defective row address stored in the redundancy ROM, a row predecoder that decodes the row address output to the address bus, and an output from the row predecoder A main decoder for decoding the row pre-decoded signal, a first latch circuit for latching the decode signal of the main decoder for each row, and a word line based on the decode signal latched by the first latch circuit. A memory cell array in which a row of memory cells is selected by being selected, a spare row cell for replacing a defective row in the memory cell array, a spare row decoder for selecting the spare row cell, and a decode output from the spare row decoder Signals for each spare row And a second latch circuit in the spare row decoder that activates the row predecoder when the address match is detected by the comparator when erasing stored data. And a logic circuit for controlling a spare row cell, the main row decoder including first bias means for applying a ground potential to a word line connected to a defective cell in the memory cell array when erasing stored data. The row decoder is provided with a second bias means for applying a ground potential to an unused spare word line when erasing stored data. 14. The first biasing means comprises the first biasing means.
A first level shifter that shifts an output signal of the latch circuit of the first level shifter to a level between a power supply potential and a bias potential, and a first level shifter that shifts an output signal of the first level shifter to a level between a first voltage and the bias potential. 2 level shifter,
The second bias means includes a first buffer for outputting the level of the output signal of the second level shifter at a level between the second potential and the bias potential, and the second bias means. Of the third level shifter for shifting the output signal of the third level shifter to a level between the power supply potential and the bias potential, and the output signal of the third level shifter for the first level shifter.
Fourth voltage level shifts to the level of the bias potential and the bias potential
And a second buffer for outputting the level of the output signal of the fourth level shifter at a level between the second potential and the bias potential, and outputting the bias potential at the time of reading and writing. The ground potential is a negative potential at the time of erasing, the first potential is a power source potential at the time of reading and erasing, and the high potential for writing at the time of writing, and the second potential is a power source potential at the time of reading and the above writing time at the time of writing. 14. The non-volatile semiconductor memory device according to claim 13, wherein the non-volatile semiconductor memory device has a high potential for use with a ground potential during erasing.